CFD reliability issues in analysis of naturally ventilated buildings

The potential for error when using computational fluid dynamics (CFD) for investigating internal building airflows continues to be a critical issue in building simulation analysis. This topic is assessed in the current paper by examining the ability of a proprietary CFD code to simulate buoyancy and forced airflow regimes, typical of a naturally ventilated building. This issue is motivated by an ongoing research project, aimed at examining the relationship between external microclimate and internal building comfort, where CFD constitutes a major analytical tool. Two experimental case studies from the literature are employed as benchmarks by which the CFD code is assessed. First, the Cheesewright study, which is based on pure buoyancy cavity flow, is considered. Results from this investigation indicate that structured meshes are not only less dependent on mesh density, but also give consistent convergence and accuracy when coupled with the k-? and k- ? turbulence models. Second, the Neilson study, which is based on forced airflow data, gave broadly similar findings, however, slightly better prediction of peak air velocities were observed when the k-? turbulence model was used. The paper concludes by discussing the relevant measures that must be considered when applying CFD methods to a full-scale naturally ventilated building.